PANW NEWS
UrgentXSIAMCISA adds PHP2Shell (CVE-2026-109618, CVE-2026-109619) to KEV; federal agencies must patch by Thursday, Oct 29UrgentXSIAMPHP2Shell (CVE-2026-109618, CVE-2026-109619) now confirmed exploited in the wildAgentiXCortex Cloud SaaS Security launches, with Cortex AgentiX handling responsePrisma"Securing the AI Systems of Tomorrow": Palo Alto's blueprint for securing Claude across the enterprisePrisma2026 SecureIQLab benchmarks name Palo Alto the sole Leader in NGFW (93.0%) and SSE (94.5%)PartnersNikesh Arora's FY2026 pay fell 19.5% to $80.3M; FY2027 PSU payout cap cut to 3xCompeteWiz: PHP2Shell (CVE-2026-109618, CVE-2026-109619) exploitation attempts and remediation trendsCompeteGoogle SecOps "What's New" (Oct 19-25): case reopen control, GTI Alerts Connector, outer joins in rulesCompeteDatadog dd-trace-py ReDoS denial-of-service bug, CVE-2026-116588 (CVSS 4.0 8.7)CompeteSplunk: Agent Observability alerts now flow into Enterprise Security for AI-driven containmentXSIAMNikesh Arora calls for "hybrid patching" and a SOC rebuilt around SOAR, then agentsPrismaPrisma Browser for Microsoft Intune is available on the Apple App Store and Google PlayPrismaPalo Alto (Idira) named a Leader in the 2026 Gartner Magic Quadrant for Privileged Access Management, for the eighth timePAN-OSCVE-2026-111497 (Brotli out-of-bounds write) affects PAN-OS 11.2 and 12.x with Brotli inspection and decryption on (9.2 CRITICAL)CompeteSentinelOne and Moro Hub bring sovereign AI security to the UAECompeteWiz: "The Pwn Request Reckoning" maps the GitHub Actions attacks that still work after GitHub's 2026 hardeningCompeteDatadog Security Labs: "Code security for new coders" guide for AI coding-agent usersCompeteSplunk SVD-2026-1013 (PHP2Shell, CVE-2026-109618 and CVE-2026-109619): Smartlook now listed as RemediedXSIAMCortex blog: mitigating PHP2Shell (CVE-2026-109618, CVE-2026-109619) with the Cortex platformXSIAMCorrection: Unit 42 retracts European government PHP2Shell (CVE-2026-109618, CVE-2026-109619) casePartnersCortex XDR One launches in IndiaCompeteWiz: 'Broccoli2Shell' Brotli decoder RCE (CVE-2026-111497) hits Node.js, Bun, Envoy, IstioCompeteWiz: 'Gheist' campaign abuses stolen GitHub tokens to dump CI secretsCompeteMicrosoft: Unified Security Operations in Defender (ISOC) runs SOC without a Sentinel workspace (preview)
This week
PAN Portfolio
HomeWhat's HotAI Conversation
Search What's Hot This Week Selling Palo Alto Networks into AI Data Centers

PANW News

Competition

63 stories with this tag. A card opens the full write-up. Search covers every tag.

Oct 26–Nov 1, 2026

WizOct 25, 2026Added Oct 26

Wiz: 'Broccoli2Shell' Brotli decoder RCE (CVE-2026-111497) hits Node.js, Bun, Envoy, Istio

Wiz (Oct 25) published guidance on an out-of-bounds write in Google's Brotli decoder affecting Node.js, Bun, Envoy, Istio, Python Brotli/brotlicffi, and Go andybalholm/brotli; researchers showed RCE against unpatched Node.js, and Wiz says 91% of cloud environments have vulnerable resources. The Hacker News reported no in-the-wild exploitation as of Oct 22. Vulnerable: libbrotli 1.0.3-1.2.0; Node.js below 22.23.4, 24.22.1, 26.12.1; Bun 1.0.21-1.4.3; Envoy 1.36.11, 1.37.7, 1.38.5, 1.39.2; Istio 1.29.0-1.29.8, 1.30.0-1.30.5, 1.31.0-1.31.1; andybalholm/brotli below v1.2.7. Fix: upgrade (Brotli 1.0.10/1.1.1/1.2.1, Python Brotli 1.2.1, brotlicffi 1.2.1.0, Node.js 22.23.4/24.22.1/26.12.1, Bun 1.4.4, Envoy 1.36.12/1.37.8/1.38.6/1.39.3/1.40.0, Istio 1.29.9/1.30.6/1.31.2, andybalholm/brotli v1.2.7), rebuild images, and restart; if you can't patch, disable Brotli decompression or reject Brotli-encoded requests.

Source · wiz.io

WizOct 25, 2026Added Oct 26

Wiz: 'Gheist' campaign abuses stolen GitHub tokens to dump CI secrets

Wiz (Oct 25) reports an automated campaign that planted sync-check workflows on 13,500+ branches across 850+ compromised accounts, dumping repo and environment secrets into logs in base32 to evade masking. GitHub's approval hold blocks runs in public repos, but they succeed in private repos and on GitHub Enterprise Server; some enterprise victims then saw AWS Secrets Manager enumeration. Check: branches matching ^ci-sync-[0-9a-z]{6}$ adding .github/workflows/sync-check.yml ('ci: sync config'); runs on ci-sync-* branches deleted within a minute; user agent 'gheist'; OIDC role assumptions with a ci-sync- subject; runner files .gdmp_pat, .gdmp_patg, .gdmp_sweep. Fix: rotate all exposed secrets, revoke the pushing tokens and check owners for infostealers, treat affected self-hosted runners as compromised, don't approve held sync-check runs, and scope OIDC trust to protected branches.

Source · wiz.io

WizOct 26, 2026Added Oct 27

Wiz: "The Pwn Request Reckoning" maps the GitHub Actions attacks that still work after GitHub's 2026 hardening

Wiz reviews GitHub's 2026 fixes: actions/checkout v7 (Jun 18), read-only cache tokens (Jun 26), a checkout backport to v2.8.0, v3.7.0, v4.4.0, v5.1.0, and v6.1.0 (Jul 20), Workflow Execution Protections GA (Sep 17), and a default pull_request_target block on public repos from Dec 2. It then lists what still works: SHA-pinned checkouts from before Jul 20, git clone or gh pr checkout bypasses, issue_comment and repository_dispatch triggers, script injection through ${{ }}, artifact poisoning, non-ephemeral self-hosted runners, and private or internal repos. Wiz pitches Wiz Code and includes a grep audit (example only).

Source · wiz.io

GoogleOct 26, 2026Added Oct 28New

Google SecOps "What's New" (Oct 19-25): case reopen control, GTI Alerts Connector, outer joins in rules

Highlights: a setting that controls whether closed cases reopen when new alerts arrive with the same grouping ID; a new Google Threat Intelligence (GTI) Alerts Connector in the SOAR marketplace; outer joins and anti-joins in detection rules; GTI_IOC feed ingestion criteria (confidence of 80 or higher, a malicious verdict, or a known threat association); and a GitOps workflow for building SOAR integrations, with Python 3.11 as the default.

Source · security.googlecloudcommunity.com

DatadogOct 27, 2026Added Oct 28New

Datadog dd-trace-py ReDoS denial-of-service bug, CVE-2026-116588 (CVSS 4.0 8.7)

The regex that obfuscates URL query strings in Datadog's Python tracer can burn CPU on attacker-controlled input and stall requests. It's on by default for affected HTTP integrations. Affected: ddtrace 1.5.0 through 3.19.8 and 4.0.0 through 4.15.5. Fixed in 3.19.9 and 4.15.6. Workaround: set DD_TRACE_OBFUSCATION_QUERY_STRING_REGEXP and DD_TRACE_OBFUSCATION_QUERY_STRING_PATTERN to the fixed regex in the advisory, then restart.

Source · github.com

Oct 19–25, 2026

WizOct 19, 2026Added Oct 19

Wiz CIRT details real AI-agent-driven attacks across cloud, Kubernetes, CI/CD, and identity

Wiz (Oct 19) says AI-driven attacks are rising; in several intrusions an AI agent ran the entire operation: exploiting an exposed Kafka UI, stealing Kubernetes service account tokens, minting GCP credentials, abusing AWS SSM, creating an IdP admin, and weaponizing CI/CD and GitOps. The same source IP scanned 600+ other Wiz customer environments. Wiz's first recommendation: unify logs across all platforms.

Source · wiz.io

SplunkOct 13, 2026Added Oct 19

Splunk: Cisco Deep Time Series Model GA in AI Toolkit; Log and Graph Reasoning previewed; advisories Oct 21

Splunk (Oct 13) made the Cisco Deep Time Series Model (CDTSM) generally available in Splunk AI Toolkit and previewed Log Reasoning and Graph Reasoning foundation models; customer Mitiga reported about 95% fewer false-positive alerts. Separately, Splunk will disclose fixes on Oct 21 for Enterprise and Universal Forwarder 10.4.4, 10.2.8, 10.0.11, and 9.4.16 and SOAR 8.7.0 (SVD-2026-1006).

Source · splunk.com

MicrosoftOct 19, 2026Added Oct 20

Microsoft extends Project Perception to secure the 'agentic insider'

On Oct 19 Microsoft extended Project Perception to govern internal AI agents: discovery of 35 more local agent types (60 total, including Claude Code, OpenAI Codex, GitHub Copilot), new Entra Agent ID identity types, Dynamic Permissions (task-scoped, time-bound), Dynamic Agent Risk, Dynamic Conditional Access for agents with graduated containment, and Guardian Agents with Playbooks. Project Perception and Agent 365 share a unified registry.

Source · microsoft.com

WizOct 19, 2026Added Oct 20

Wiz detects npm supply chain zero-days at package publication

Wiz (Oct 19) described a pipeline that analyzes each new npm package version at publication using static analysis, an AI classifier, and sandboxed runtime analysis with the Wiz Sensor, then publishes IOCs to Wiz Defend and code and cloud scanning before a CVE exists. Wiz says notable supply chain attacks rose from about 10% of significant incidents in H2 2025 to 25% in H1 2026.

Source · wiz.io

DatadogOct 19, 2026Added Oct 20

Datadog upgraded by Bernstein on AI-driven observability demand

CNBC reports Bernstein upgraded Datadog to Outperform on Oct 19 and raised its price target to $350 from $237, citing alternative data showing enterprise token consumption at the largest AI labs grew at least 50% quarter over quarter; Bernstein flagged open-source and in-house replacement as the main risk. Shares were up about 82% year to date. Datadog reports Q3 results Nov 4.

Source · cnbc.com

SplunkOct 19, 2026Added Oct 20

Splunk Edge Processor, Ingest Processor, and AI Assistant achieve FedRAMP High

Splunk (Oct 19) said Edge Processor, Ingest Processor, and Splunk AI Assistant achieved FedRAMP High for Splunk Cloud Platform customers in AWS GovCloud (US), and SPL2 is now available across the platform. Separately, Splunk will publish advisories on Oct 21 for fixes in Splunk Enterprise and Universal Forwarder 10.4.4, 10.2.8, 10.0.11, 9.4.16 and SOAR 8.7.0.

Source · splunk.com

SentinelOneOct 21, 2026Added Oct 21

SentinelOne launches Singularity ASOC autonomous SOC suite at OneCon

On Oct 21 at OneCon, SentinelOne unveiled Singularity ASOC (Autonomous SOC). It bundles Purple Agentic Investigation, Hyperautomation, AI Data Pipelines (50+ connectors, including Palo Alto Prisma, Microsoft Defender XDR, and Wiz; OpenTelemetry and generic API collection listed as coming), AI SIEM with incident and case management, and threat-intel/IOC management. It is paid for with Singularity Credits, with analyst-set consumption guardrails. SentinelOne says Purple Agentic Investigation has completed 400,000+ investigations and saved about 332,000 analyst hours. (SentinelOne's own press-release URL returned 404 when checked on Oct 21; source is WebProNews coverage.)

Source · webpronews.com

WizOct 20, 2026Added Oct 21

Wiz CIRT: phishing-to-LLMjacking attacks hijack cloud billing accounts for AI usage

Wiz CIRT (Oct 20) describes GCP-focused intrusions in which Google Search ads for terms like 'google cloud' led to fake login pages that captured credentials and MFA codes. Attackers then enabled the Gemini API in low-usage projects, created service accounts with Gemini API keys and high-privilege roles such as roles/ml.admin, and linked attacker-controlled GCP projects to the victim's billing account. Victims paid for hundreds of thousands of dollars in AI API charges without seeing the prompts or outputs.

Source · wiz.io

GoogleOct 19, 2026Added Oct 21

Google SecOps extends SOAR-to-Chronicle API migration deadline to Mar 31, 2027

Google's weekly 'What's New in Google SecOps' (Oct 19, covering Oct 12-18) says the deadline for migrating SOAR APIs to the Chronicle API moved from Nov 30, 2026 to Mar 31, 2027. Google also published parity docs for moving MSSPs off the legacy Backstory Partner API, Applied Threat Intelligence curated rules now use the GTI_IOC feed, and Google TI added a macOS Mirage Sandbox. The roundup's top item was Gemini 4 Argon's release.

Source · security.googlecloudcommunity.com

CrowdStrikeOct 21, 2026Added Oct 22

CrowdStrike: Falcon Guardian to integrate with NVIDIA OpenShell, with Sentry planned

CrowdStrike is integrating Falcon Guardian (AI Detection and Response) with the NVIDIA OpenShell runtime and plans to integrate with NVIDIA Sentry on BlueField-4 for in-silicon enforcement. Autodesk is quoted. Separately, CrowdStrike published PHP2Shell coverage: post-exploitation detection, Exposure Management checks, and a Next-Gen SIEM correlation template.

Source · crowdstrike.com

WizOct 21, 2026Added Oct 22

Wiz Research discovers and discloses PHP2Shell

Wiz says every PHP version back to 5.4 is affected by at least one of the PHP2Shell bugs and showed pre-auth RCE on WordPress, Drupal, and Joomla. It used Wiz Atlas, Gemini 4 Argon, and GPT-5.6 Sol to find them. Wiz Defend and the Wiz Sensor detect exploitation, with an automatic response policy on Sensor version 1.0.12768.

Source · wiz.io

SplunkOct 21, 2026Added Oct 22

Splunk ships its October patch batch for Enterprise and SOAR

SVD-2026-1007 (Oct 21) fixes multiple CVEs in Splunk Enterprise 10.4.4, 10.2.8, 10.0.11, and 9.4.16, including RCE through Splunk Web Manager XML, authorization bypass, and SPL injection in Secure Gateway. SecurityWeek counts 16 Enterprise CVEs (including critical CVE-2026-76299 and CVE-2026-76302) and 15 in SOAR 8.7.0 (including critical CVE-2026-76531 and CVE-2026-76407). The Universal Forwarder was dropped from the batch.

Source · advisory.splunk.com

CrowdStrikeOct 21, 2026Added Oct 23

CrowdStrike details four new agentic-AI threat actors

CrowdStrike's Oct 21 Counter Adversary Operations report describes four previously unreported actors using AI agent harnesses; one ran a Claude Code supervisor/executor setup derived from PentestGPT against at least 15 Argentine healthcare organizations. Per Cybersecurity Dive, the report counts three times as many agentic intrusions in H2 2026 as H1 and says token budgets, not engineering, are the main barrier to scale. CrowdStrike assesses agentic intrusions will almost certainly increase. Defender guidance is in the gated full report.

Source · crowdstrike.com

SentinelOneOct 21, 2026Added Oct 23

SentinelOne Prompt Security joins Lenovo ThinkShield

SentinelOne announced on Oct 21 that Prompt Security is now part of Lenovo's ThinkShield portfolio, discovering sanctioned and shadow AI use, stopping sensitive data from reaching unauthorized AI apps, and enforcing policy in real time, building on SentinelOne EDR already in ThinkShield.

Source · sentinelone.com

WizOct 22, 2026Added Oct 23

Wiz: ARTEX AI pentest agent used against Okta, Entra, and Cognito, overlapping with Storm-0940

Wiz reported on Oct 22 that ARTEX, an open-source autonomous pentest framework (closed-source since Oct 8 after 12,000+ forks), was used throughout October against IT and software vendors: scraping JavaScript source maps and OpenAPI docs, unsigned JWT (alg: none) SSO bypass attempts, burst password spraying of Okta and Entra, and account-recovery abuse, rotating IPs and user agents when blocked. Wiz notes overlap with Microsoft's Storm-0940 report, which Microsoft links to a supply-chain attack using stolen vendor credentials against downstream Microsoft 365 tenants. Wiz shares IOCs, example log patterns, and detection rule names.

Source · wiz.io

DatadogOct 22, 2026Added Oct 23

Datadog adds AI cost features to Agent Observability

Datadog announced on Oct 22 a Cost page in Agent Observability that ties estimated AI spend to applications, traces, models, providers, and tags, plus Cost Savings Opportunities (for example uncached prompt prefixes or full history re-sent each turn) with a Fix with Bits option. Agent Observability is free for up to 40,000 LLM spans a month.

Source · datadoghq.com

SplunkOct 22, 2026Added Oct 23

Splunk demos a five-agent attack swarm in Attack Range; PHP2Shell advisory SVD-2026-1013

Splunk's Threat Research Team showed on Oct 22 five cooperating attack agents reaching Domain Admin, DCSync, and Linux root in 3 minutes 47 seconds in a lab; password spraying (T1110.003) was the top technique, and the defensive framework shown is a personal experimental project, not a Splunk product. Separately, SVD-2026-1013 (Oct 22) says no Splunk products are currently known to be affected by PHP2Shell (CVE-2026-109618, CVE-2026-109619); Splunk OnCall and AppDynamics SaaS are under investigation.

Source · splunk.com

CrowdStrikeOct 23, 2026Added Oct 24

CrowdStrike says PHP2Shell is hard to weaponize; reports no successful exploitation

CrowdStrike Counter Adversary Operations (Oct 23) says most public PHP2Shell PoCs only crash PHP, reliable RCE must be tuned per PHP version, SAPI, app, and image, and it has seen only probing and failed attempts, with no successful exploitation in the wild or in honeypots. It recommends patching, rebuilding containers, restarting workers, disabling upload progress (which doesn't mitigate CVE-2026-109618), turning off file_uploads where unneeded, and limiting request bodies at the proxy or WAF.

Source · crowdstrike.com

MicrosoftOct 23, 2026Added Oct 24

Microsoft details the LiteLLM 1.107.0 supply-chain compromise

Microsoft (Oct 23) analyzed malicious LiteLLM 1.107.0, published Oct 21 on PyPI and official container images. It plants cook.pth, cook.py, and sitecustomize.py so starting Python runs a credential stealer even without importing LiteLLM, targeting AI-tool, cloud, Git, SSH, and browser credentials and LiteLLM proxy database data, exfiltrating to www.litellm[.]dev. Guidance: find every copy (environments, caches, mirrors, lock files, images), don't run python or pip in a suspect environment, rebuild from verified artifacts, rotate credentials from a clean machine, hash-pin dependencies, and restrict egress. Microsoft provides Defender detections and hunting queries.

Source · microsoft.com

DatadogOct 23, 2026Added Oct 24

Datadog Agent 7.84.6 blocks compromised eMudhra intermediates (CVE-2026-115557)

Datadog Agent 7.84.6 (Oct 23) makes its embedded Python reject TLS chains containing compromised eMudhra intermediate CA keys regardless of trust store, backporting the CPython fix for CVE-2026-115557; on Windows, Python checks no longer trust those chains via the Windows certificate store. Fix: upgrade Agents and Cluster Agents to 7.84.6; separately, upgrade pip to 26.2.2+ and certifi to 2026.10.19.

Source · github.com

SentinelOneOct 21, 2026Added Oct 25New

SentinelOne: 'Stopping PHP2Shell' (CVE-2026-109618, CVE-2026-109619) post-exploitation coverage

SentinelOne says its Singularity agent and AI SIEM content detect post-exploitation behavior after web-server flaws (shell commands, web shells, payload downloads, outbound connections) and Purple AI can hunt for it, while stating patching is the only complete fix. Guidance: patch and restart Apache and PHP-FPM, inventory bundled PHP, isolate PHP 8.1 and older, rotate secrets a PHP worker could hold.

Source · sentinelone.com

Oct 12–18, 2026

MicrosoftOct 16, 2026Added Oct 17

Microsoft and NVIDIA publish open-weight model risk guidance

Risk Management for Open-Weight Models in Enterprise introduces a five-category taxonomy (Cybersecurity, Data Protection, Content, Human-AI Interaction, Bias) grounded in ISO and NIST AI RMF, mapped to evaluations, pointing to Microsoft Foundry. Covers the model layer; runtime, harness, and application layers come later.

Source · microsoft.com

SplunkOct 14, 2026Added Oct 17

Splunk adds MCP Server, Agent Skills, splunkctl, and Agentic SDK

Splunk released tools that let customers' own agents query and operate Splunk (MCP Server, 30+ Agent Skills, open-source splunkctl CLI, Agentic Splunk SDK), with an app-dev agent harness in controlled availability and human approval for consequential actions. Separately, SVD-2026-1006 pre-announces Oct 21 advisories with fixes in Enterprise 10.4.4, 10.2.8, 10.0.11, 9.4.16 and SOAR 8.7.0.

Source · splunk.com

CrowdStrikeOct 16, 2026Added Oct 18New

CrowdStrike Falcon Guardian adds Claude Code hooks to enforce AI coding-agent policy

Oct 16 demo: Falcon Guardian checks Claude Code prompts and tool calls against central policy via hooks (UserPromptSubmit, PreToolUse, PostToolUse) deployed through a managed settings file developers cannot override. It blocks prompts with sensitive data, redacts sensitive content from files Claude reads, stops commands before they run, and logs each prompt and tool call by user.

Source · securitysenses.com

SentinelOneOct 16, 2026Added Oct 18New

SentinelOne completes $1.25B convertible notes offering; 8-K cites potential acquisitions

On Oct 16 SentinelOne completed its $1.25B offering of 0.00% convertible senior notes due 2031, including full exercise of the $187.5M option. It used about $151.4M for capped calls and about $150M to repurchase shares; the remainder may fund general corporate purposes, additional buybacks, and potential acquisitions and strategic investments. OneCon runs Oct 20 to 22 in Las Vegas.

Source · sec.gov

MicrosoftOct 14, 2026Added Oct 18New

Microsoft details and helps disrupt phishing-infrastructure ecosystem (Storm-3119, Storm-2990)

Oct 14: Microsoft Threat Intelligence and the Digital Crimes Unit described Storm-3119 (YAM Store, an initial access broker abusing hijacked M365/Azure for Education tenants) and Storm-2990 phishing kits including Iron Man System. Microsoft says the ecosystem sent over 1 billion phishing emails and SMS messages to more than 75,000 organizations since January 2026, using techniques including device code phishing; DCU facilitated a disruption with law enforcement.

Source · microsoft.com